【作者】 王凌谊；

【导师】 侯世英；

【作者基本信息】 重庆大学 ， 电气工程， 2007， 硕士

【摘要】 随着国民经济的快速发展,各个行业对电能质量的要求不断提高。电力系统无功的合理分布是保证电压质量和降低网损的前提条件。电力系统的无功优化是一个复杂的非线性优化问题,有效合理的无功优化不仅能够保证电压质量和降低网损,而且对系统安全性和经济性有着重要的意义。本文首先介绍了电力系统无功优化问题研究的内容和现状。在综述了应用于电力系统无功优化问题求解的各种优化算法和分析了各种优化算法的优缺点和适用范围的基础上,以系统网损最小为目标,同时满足潮流约束和安全约束,建立了的无功优化模型。推导了目标函数和约束条件的梯度公式,为求解无功优化问题的编程实现提供了基础。通过比较经典算法中求解非线性规划问题的各类算法的优缺点,选用了具有代表性的序列二次规划法和原对偶内点法来求解无功优化问题,设计了求解无功优化问题的步骤,用MATLAB语言编写了具体的优化程序。仿真验证表明,虽然这两种算法收敛速度快,但是对模型和起始点的要求高,而且一般只能搜索到局部最优解。应用遗传算法求解无功优化问题可以克服经典算法对模型和起始点的依赖,并能搜索到全局最优解。本文将标准遗传算法和求解约束问题数值优化的遗传算法(多种群遗传算法)应用于无功优化问题的求解,针对两种算法各自的一些不足对算法作了改进,设计了求解程序。仿真验证表明,虽然遗传算法对求解问题没有什么限制,能求得全局最优解,但存在局部搜索能力不强和容易早熟的问题。结合经典算法和遗传算法各自的优点,本文提出了三种混合遗传算法:(1)遗传算法和序列二次规划法的混合遗传算法。先用遗传算法进行解域搜索,得到的一组中间结果作为序列二次规划法的初始值求解最优解。这样的混合方式不仅解决了序列二次规划法的初值问题,而且后期采用收敛速度快的序列二次规划法搜索,提高了整体算法的计算速度和收敛性。(2)遗传算法和原对偶内点法的混合遗传算法。先用原对偶内点法求得一组次优解,再以这一组次优解作为遗传算法的初始种群,求解到最终的全局最优解。这种混合方式不仅利用了原对偶内点法计算速度快且不随网络规模的增大而减小的优势,弥补了遗传算法初期计算潮流速度慢的缺点,提高了整体算法的计算速度,而且保证了最终解的全局性。(3)混合多种群遗传算法。将多种群算法中的参照群体的遗传操作改为使用序列二次规划法产生一个新的个体。混合算法从一定程度上减少求解无功优化问题时个体的测试、组合和适应值计算,提高算法的搜索效率。最后,以IEEE-14节点系统为对象,分别采用了各种算法进行优化计算,通过仿真结果和已有文献的结果的对比分析,验证了本文所提出方法的正确性和有效性。更多还原

【Abstract】 With the development of national economy, the demands of power supply quality from all kinds of industries are increased. Rational distribution of reactive power in system is the prior condition which can ensure voltage quality and reduce the network loss. Reactive power optimization is a kind of complicated nonlinear optimization problem. Efficent and reasonable reactive power optimization not only can ensure voltage quality and reduce the net loss, but also be very important for security and economics of power system.In this paper, the study contents and present status of reactive power optimization are represented. After researched kinds of methods for reactive power optimization, the advantages and disadvantages and application of these methods are analyzed. A mathematical model for reactive power optimization is established to obtain minimization of network loss with satisfying constraints of power flow and security, and the gradients of constraints and object which are the basement of using classic algorithms for promble solving are derived.After compared the merits and faults of optimization methods in classic algorithms for nonlinear optimization problem, the Sequential Quadratic Programming (SQP) and Primal-Dual Interior Point (PDIP) methods which are representative are adopted for reactive power optimization. The principles of two methods are introduced, then the steps for reactive power optimization are designed and the optimization programs for reactive power optimization are compiled using MATLAB languages. The simulation results prove that though the convergence speed for the two kinds of methods is fast, the methods are indepent on the maths model and initial point and usually can reach the local optimum result.Using Genetic Algorithm (GA) to solve the reactive power optimization problem not only can avoid the faults of classic algorithms such as depending on the model, initial start points and so on, but also have the ability to search the global optimum result. In this paper, two kinds of GA which are standard GA and Genetic algorithm for numerical optimization of constrained problems (a kind of multi-pop synergism evlution GA) are employed to solve the reactive power optimization. Some faults are improved and counterpart programs are compiled. The simulation results prove that though there are hardly limits for GA sovling the problem and the GA can effectively get the global optimum result, there are some defaults that GA is lack of ability to search local domain quickly and easy to be premature.Combined the merits of classic algorithm and GA, three kinds of hybrid GA are composed.(1) The hybrid GA with GA and SQP. In this kind of hybrid GA, GA is firstly used for searching domain and obtains a group of middle results, and then SQP is employed for searching finally optimum result using this group of middle results as its initial point. The hybrid GA can not only solve the problem of selecting initial point for SQP, but also increase the convergence speed.(2) The hybrid GA with GA and PDIP. PDIP is firstly used for searching a group of suboptimum results, and then GA is employed for searching the finally optimum result. The hybrid GA can not only take advantage of computiton speed of PDIP to remedy the computition speed of GA computing the power flow in the beginning, but also ensure the global result can be obtained.(3) The hybrid multi-pop GA. This kind of hybrid GA uses SQP to replace the genetic operation in birthing reference population is represented. The hybrid multi-population GA can partly decrease the number of computing the adaptive values and develop the search efficiency.At last, all the proposed methods are applied to IEEE 14-bus system. The numerical simulation results compared with exsited literature demonstrate the validity and efficiency of these algorithms.更多还原